CX thought leader shares how CX can use AI to earn a seat at the table
Jim Tincher explains how CX leaders can use AI to connect customer signals with revenue, prioritize journey fixes, and show executives the financial value of CX.
Jim Tincher
Founder and CEO Heart of the Customer
AI Boundaries: Apply AI to analyze customer data, while keeping relationship decisions and service recovery human.
Revenue Focus: Rank journey improvements by revenue risk, not complaint volume, to identify silent accounts likely to shrink.
Proof Matters: Trace every AI-generated quote and number to its source, because plausible errors can distort strategic decisions.
Break Silos: Connect operational, emotional, behavioral, and financial data to reveal where customer experience changes affect growth.
Earn Credibility: Show executives how customer emotions predict behavior and revenue instead of presenting satisfaction metrics alone.
Jim Tincher is a CX thought leader and researcher. He's also the founder of Heart of the Customer, a boutique E2E customer experience firm supporting complex organizations.
We sat down with Jim to learn what he's seeing in his research. He went deep on using AI to enhance customer journeys.
A career in CX and research
I'm Jim Tincher, founder and CEO of Heart of the Customer. We've been doing this since 2012.
Continue Reading for Free
Create a free account to finish this article, plus get ongoing access to timely insights and practical resources.
Have an account? Log In
I don't run a customer experience department — I advise the executives who carry the growth number: Chief commercial officers, VPs of sales and operations, CEOs. We work across B2B; our deepest work is in manufacturing: BASF, Dow, Milliken, Xylem, Legrand, Vermeer.
I also lead research analysis for the Advanced Manufacturing CX Consortium — seventeen manufacturers who together survey nearly 10,000 of their customers a year to determine what makes a customer grow with a supplier.
Why AI should be used on data, not relationships
With the advancement of AI, we changed how we do analysis, not our customer relationships. — a deliberate distinction.
Last year, I interviewed 34 C-Suite executives to learn what mattered most to them, and the answer (unsurprisingly) was "growth." All 34 agreed, so we use that lens for everything.
Our benchmark found the top three predictors of customer growth are all relationship signals: Feeling valued, the account relationship, and the buying process; not ease of doing business, satisfaction, or other typical CX measures. This makes me cautious about automating areas where customers decide if their business matters to you. Stripping out the human element may remove the very signals customers use to judge whether to grow with you.
So, we apply AI to analysis instead — analyzing conversations, open-ended comments, and other communications. Manually reading, coding, and finding patterns across hundreds of conversations is slow and limits our capacity. But with Claude and a transcription tool, we can search for patterns across an entire engagement's history, rather than one touchpoint at a time.
AI enabled us to search across 800+ B2B interviews in our data, finding structure we couldn't see before. We then tested the four-tier hierarchy, which predicts how much of a customer's spend you'll get, against twenty-one separate client datasets. I could not have done that analysis by hand at that scale.
A workflow that provides a ranked list of steps to fix in the customer journey
Here's a workflow that ends in a ranked list of journey steps to fix, ordered by the revenue at stake.
It starts with conversations, not surveys. We interview a client's customers directly — the buyer, the plant manager, the executive who signs the contract — and record and transcribe those conversations through Fireflies.
That's where the old constraint lived. An analyst reads what they have time to read, and that ceiling shapes every conclusion. You sample, you find a pattern in the sample, and you hope the sample was representative.
Next, the analyst assembles initial findings - the story, the key themes.
Then, we turn to Claude. Claude reads every transcript in full. It pulls out what customers said, verbatim, with the source attached: who said it, which conversation, where in the conversation. That provenance matters more than the speed. When a client challenges a finding, we don't defend it. We show them the quote.
Then, we perform the analysis, and this part is not automated.
Is this customer still discussing basic functionality? Or are they discussing whether they can count on our client when things get complicated? Or whether they feel valued? That gives us a position, not a score.
Then, we map it across the journey. A curve emerges — confidence builds at this step, breaks at that one. Most clients have never seen their relationship that way. They've seen one satisfaction number for the whole relationship, which obscures the exact place damage occurs.
We connect breaks to accounts and accounts to money. Not "customers are frustrated at this step." Instead: confidence drops here, these eleven accounts sit below that line, and here's how their spending has changed over the last four quarters.
That ordering is the point. Most companies prioritize by complaint volume — which is nearly random, because accounts quietly planning their exit are usually those not complaining.
Finally, we use Reduct to transform the critical quotes into videos, sharing the customer's voice.
The division of labor is key. AI removed a reading constraint. It did not affect judgment — deciding what to ask, recognizing which patterns matter, knowing when a customer is telling you something they don't know they're telling you. This represents fourteen years of work, and it does not automate.
Why service recovery must remain human
So, I draw the line here: AI on the data, humans when customers decide if they matter to you. AI makes no decisions — CX or otherwise — though I use it to inform my decisions.
So, AI is best used for
- Finding patterns across conversations and open-ended comments.
- Figuring out which issues actually cost money — most companies prioritize by complaint volume, a nearly random method.
- Detecting accounts that have gone quiet, those with narrowing order breadth, or those whose language has shifted. Humans miss this signal because no one's job is to watch for it.
The most important place to keep things human is service recovery. Our data is blunt here — customers completely satisfied with an issue's resolution shrink their spend 1% of the time. Customers not satisfied with the resolution shrink their spend 35% of the time. And "somewhat satisfied" customers behave more like dissatisfied customers than satisfied ones. Partial fixes count as no fix.
This last finding is the real argument. Automation excels at producing adequate outcomes at scale. Adequate is exactly where growth dies.
For escalation, I'd split it: let AI find the account at risk, then a human calls. Detection is a data problem. Recovery is a relationship problem. And I'm most skeptical of automated personalization, because B2B customers can tell, and personalization that reads as generated signals the opposite of its intent.
Why AI analysis requires strict guardrails
So, AI excels at finding patterns. But that includes illusory ones.
AI invents specificity. Small, plausible findings: a number in the right range, a quote that sounds right. And if you seek support for something you already believe, you'll find it.
So, AI excels at finding patterns. But that includes illusory ones… Every number and quote we publish must trace back to a specific source, and when something doesn’t reconcile, we flag it rather than smooth it over. AI just raised both our ceiling and the level of discipline required.
So, we've become stricter, not looser: Every number and quote we publish must trace back to a specific source, and when something doesn't reconcile, we flag it rather than smooth it over. AI just raised both our ceiling and the level of discipline required.
How AI can remove organizational silos
Everyone redesigns the customer-facing layer — chatbots, deflection, automated outreach. Almost nobody redesigns the layer underneath, where the company's knowledge sits in pieces.
Operations knows your on-time delivery. Sales knows which accounts stopped ordering a product line. Digital knows who orders, abandons orders, or gets stuck trying. Finance knows what happened to margin. CX knows how customers feel. Five groups, five datasets, five systems. Nobody holds all five.
The answer to "Why did that account shrink?" exists inside your building, but it's distributed across people who don't sit in the same meeting. As I tell my clients, "You only knew what you know." What could happen if you were able to unleash all the disjointed knowledge in your organization? What would that enable?
It's not a data problem. Adding a fifth dashboard won't fix it. It's a model problem — without a chain connecting operations to emotion to behavior to revenue, more data gives you more disconnected data.
This is where AI earns its keep, and it isn't the use case getting the attention. It can hold all five of those at once. It can read emails, sit alongside operational history and order patterns, and find where they connect.
The technical part was the easy part. Getting the data is the hard part. Every time. That's why this doesn't work as an analytics project. It requires a sponsor — someone senior enough to convene the various groups who hold the data, and to challenge the groups to find out where growth is leaking across the silos of information.
Why organizations are focusing on the wrong AI metrics
Ask most companies what AI improved, and they'll mention deflection rates and response times. In other words, they'll tell you how much human contact they removed. Then they'll say satisfaction held steady.
It probably did. That's the trap.
In our benchmark, 81% of customers were satisfied. Only 27% planned to grow. Satisfaction survived. Growth didn't.
Here's what I'd do. Look at what you automated last year, then overlay that with the stages of the journey where your customers determine whether you value their business. If you automated away those critical touch points, you didn't save on costs — you inadvertently cut your revenue forecast.
How CX leaders can use AI to earn a seat at the table
As I said, I interviewed 34 C-Suite executives about their priorities, and every one named growth. Almost none of them raised customer experience as an important topic on their own — despite them knowing that I'm a customer experience thought leader, that the CXPA sponsored the meeting, and that the meeting invitation included CX.
That reads like bad news. It isn't. The disconnect isn't that your executives don't care about customers — it's that nobody has shown them how customer experience moves dollars.
Executives cared far more about what customers did than what customers said. When I asked what they watch to judge the health of the business, they named P&L trends, fulfillment metrics, on-time delivery, pipeline velocity, RFP wins. Retention. Repurchases. Sentiment measures — NPS included — almost never came up. This aligns with the XM Institute's finding that most CX programs can't demonstrate value. Why would a C-Suite member bring up CX?
One of them put it better than I can. "We have dashboards filled with lagging indicators. Lagging indicators are easy. It's finding the leading ones that's hard."
Don't skip over that, because it's your opening. Emotions and behaviors are the leading indicators. And you own the emotion data — assuming you're measuring emotions (and you should be). Nobody else in your company has that data.
So here's the advice: Stop presenting what customers feel. Start presenting what it predicts.
Stop asking for a seat at the table. You don’t get the seat by asking for it. You get it by showing your CEO that a specific operational investment changed how customers felt — and that the change moved revenue.
There's a chain running from operations to revenue. Your company pulls operational levers — delivery, communication, problem resolution. Those create emotions in your customers. Emotions change behavior. Behavior shows up as dollars. Levers, emotion, action, dollars — it's called The Growth Chain™ , and we use the LEAD acronym to make it memorable. Operations owns the first link, finance owns the last, and the two in the middle are yours. They're also the two nobody is measuring.
This is where AI can be indispensable: discovering and reporting on these connections.
Three out of four CX programs never demonstrate a connection to financial outcomes. That's the number that should worry you, because it's the number that gets your budget cut the moment margins tighten.
Where to start is smaller than you'd think. Pick two accounts — one growing, one declining. Trace the chain on both. On the growing account, work forward from the operational change. On the declining one, work backward from the revenue.
Then, add three questions to a survey you're already sending. How confident are you that we can solve your future needs? How valued do you feel as our customer? To what extent do we operate as a true partner in your business?
Two accounts and three questions. Do that this quarter and you'll walk into a room with something nobody else in your company can produce.
And stop asking for a seat at the table. You don't get the seat by asking for it. You get it by showing your CEO that a specific operational investment changed how customers felt — and that the change moved revenue.
Follow Along
You can follow Jim Tincher's work on LinkedIn. And check out Heart of the Customer.
More expert interviews to come on The CX Lead!
